Early impressions count: Category ratings established from small samples

We examined the establishment of category ratings from small samples in three studies. A Monte Carlo experiment showed that basing categories on the first five or ten numbers encountered could provide a usable way to categorize numbers drawn from flat, normal and skewed distributions. Experiment 2 in which people categorized random integers presented one by one into five bins showed that after five or ten presentations the category structure was quite similar to that obtained after 30 presentations, and Experiment 3 found a similar result when people categorized arc segments. Overall, the results show that usable category ratings can be established from relatively small samples of observations.